Dataset
The collection of data an ML model trains on — quality and size matter more than architecture.
"Garbage in, garbage out": a model trained on biased data becomes biased. Classic sets: ImageNet (14M images), Common Crawl (for LLMs), MNIST (digits for beginners).